Human body shape models play important parts in three-dimensional (3D) games or the like, as well as computer animations. In order to implement the animations of such human body models, generation of a human body model enabling shape deformation must be performed in advance. The generation of the human body model plays an important role in determining the quality of the animations.
Human body models generated according to conventional technology include a stick model in which only an initial skeleton is modeled, a surface model in which the outward appearance of a human body is represented by surface patches, and a volume model in which the human body is configured using the combination of a sphere, a cylinder, an ellipsoid, etc.
However, such models are disadvantageous in that it is difficult to realistically represent the outward appearance of the human body or to naturally deform shapes corresponding to motions. Also, they require a lot of computation time for shape deformation or require a manual operation of a user such as a professional designer.
Recently, there have been proposed a muscle simulation model in which anatomical features are incorporated, and a linearly combined interactive skinning model based on example data composed of a skeleton and a mesh structure, etc. However, despite the advantage of enabling relatively realistic deformation of a shape, such models are problematic in that, it is difficult to perform deformation in real time and to produce the model due to limitation of computational speed, and in that the precision of generated animation is determined in accordance with the precision of previously produced models and a degree of the combination of the models, and artifacts such as the ‘candy-wrapper’ effect occur on principal joints.